Skip to main content

DATA Skills (14)

Part of Role Skills Catalog | Phase 4 + Phase 6

Chains: Events → Metrics → Dictionary → Dashboard | Funnel → Experiment → Analysis → Cohort → Report | Attribution → LTV/CAC

Userflow Schema​

flowchart TD
    jaan-to-data-event-spec["data-event-spec<br>Event Spec<br>GA4 events + params + triggers"] --> data-gtm-datalayer["data-gtm-datalayer<br>GTM DataLayer ★<br>Tracking code"]
    jaan-to-data-event-spec["data-event-spec<br>Event Spec<br>GA4 events + params + triggers"] --> jaan-to-data-metric-spec["data-metric-spec<br>Metric Spec<br>Formula + caveats + segmentation"]
    jaan-to-data-metric-spec["data-metric-spec<br>Metric Spec<br>Formula + caveats + segmentation"] --> jaan-to-data-metric-dictionary["data-metric-dictionary<br>Metric Dictionary<br>Definitions + SQL + pitfalls"]
    jaan-to-data-metric-dictionary["data-metric-dictionary<br>Metric Dictionary<br>Definitions + SQL + pitfalls"] --> jaan-to-data-dashboard-spec["data-dashboard-spec<br>Dashboard Spec<br>Layout + filters + cadence"]
    jaan-to-data-funnel-review["data-funnel-review<br>Funnel Review<br>Baseline + drop-offs + hypotheses"] --> jaan-to-data-experiment-design["data-experiment-design<br>Experiment Design<br>Hypothesis + ramp/kill criteria"]
    jaan-to-data-funnel-review["data-funnel-review<br>Funnel Review<br>Baseline + drop-offs + hypotheses"] --> jaan-to-data-cohort-analyze["data-cohort-analyze<br>Cohort Analyze<br>Retention curves + churn risk"]
    jaan-to-data-experiment-design["data-experiment-design<br>Experiment Design<br>Hypothesis + ramp/kill criteria"] --> jaan-to-data-analysis-plan["data-analysis-plan<br>Analysis Plan<br>Sample sizing + decision rules"]
    jaan-to-data-analysis-plan["data-analysis-plan<br>Analysis Plan<br>Sample sizing + decision rules"] --> jaan-to-data-post-launch-report["data-post-launch-report<br>Post-Launch Report<br>Insights + chart checklist"]
    jaan-to-data-cohort-analyze["data-cohort-analyze<br>Cohort Analyze<br>Retention curves + churn risk"] --> jaan-to-data-post-launch-report["data-post-launch-report<br>Post-Launch Report<br>Insights + chart checklist"]
    jaan-to-data-post-launch-report["data-post-launch-report<br>Post-Launch Report<br>Insights + chart checklist"] -.-> jaan-to-pm-release-review["pm-release-review<br>PM: release-review"]
    jaan-to-data-attribution-plan["data-attribution-plan<br>Attribution Plan<br>UTMs + governance + limits"] --> jaan-to-data-ltv-cac-model["data-ltv-cac-model<br>LTV/CAC Model<br>Inputs/outputs + sensitivity"]
    jaan-to-data-anomaly-triage["data-anomaly-triage<br>Anomaly Triage<br>Scope + likely causes + RCA"] -.-> jaan-to-sre-incident-runbook["sre-incident-runbook<br>SRE: incident-runbook"]
    jaan-to-data-dbt-model["data-dbt-model<br>dbt Model<br>Staging/mart + tests + schema"] --> jaan-to-data-sql-query["data-sql-query<br>SQL Query<br>Ad-hoc SQL + results summary"]

    style data-gtm-datalayer fill:#e8f5e9,stroke:#4caf50
    style jaan-to-pm-release-review fill:#f0f0f0,stroke:#999
    style jaan-to-sre-incident-runbook fill:#f0f0f0,stroke:#999

Legend: Solid = internal | Dashed = cross-role exit | Gray nodes = other roles

/jaan-to:data-event-spec​

  • Logical: data-event-spec
  • Description: GA4-ready event/param spec: naming, triggers, required properties, GTM implementation notes
  • Quick Win: Yes - extends gtm-datalayer pattern
  • Key Points:
    • Events are verbs; properties add context
    • Ensure consistent naming + schema
    • Validate tracking with QA and dashboards
  • → Next: data-gtm-datalayer, data-metric-spec
  • MCP Required: GA4 (measurement alignment)
  • Input: [initiative]
  • Output: $JAAN_OUTPUTS_DIR/data/events/{slug}/event-spec.md

/jaan-to:data-metric-spec​

  • Logical: data-metric-spec
  • Description: Metric definition: formula, caveats, segmentation rules, owner, gaming prevention
  • Quick Win: Yes - simple definition
  • Key Points:
    • Define metrics precisely (numerator/denominator)
    • Include guardrails and anomaly callouts
    • Make it "actionable by default"
  • → Next: data-metric-dictionary
  • MCP Required: GA4 (dimension/metric checks)
  • Input: [metric]
  • Output: $JAAN_OUTPUTS_DIR/data/metrics/{slug}/metric-spec.md

/jaan-to:data-metric-dictionary​

  • Logical: data-metric-dictionary
  • Description: Metric definitions + SQL-like logic description, pitfalls + edge cases, example interpretations
  • Quick Win: Yes
  • Key Points:
    • Define metrics precisely (numerator/denominator)
    • Include guardrails and anomaly callouts
    • Make it "actionable by default"
  • → Next: data-dashboard-spec
  • MCP Required: None
  • Input: [metrics]
  • Output: $JAAN_OUTPUTS_DIR/data/metrics/{slug}/metric-dictionary.md

/jaan-to:data-dashboard-spec​

  • Logical: data-dashboard-spec
  • Description: Dashboard layout + sections, definitions + filters, recommended review cadence
  • Quick Win: Yes
  • Key Points:
    • Define metrics precisely (numerator/denominator)
    • Include guardrails and anomaly callouts
    • Make it "actionable by default"
  • → Next: —
  • MCP Required: None
  • Input: [kpis]
  • Output: $JAAN_OUTPUTS_DIR/data/dashboard/{slug}/dashboard-spec.md

/jaan-to:data-funnel-review​

  • Logical: data-funnel-review
  • Description: Funnel baseline + top drop-offs + segments + 3-5 hypotheses ranked by impact × confidence
  • Quick Win: No - needs GA4 MCP
  • Key Points:
    • Events are verbs; properties add context
    • Ensure consistent naming + schema
    • Validate tracking with QA and dashboards
  • → Next: data-experiment-design, data-cohort-analyze
  • MCP Required: GA4 (funnel analysis), Clarity (qualitative)
  • Input: [initiative]
  • Output: $JAAN_OUTPUTS_DIR/data/insights/{slug}/funnel-review.md

/jaan-to:data-experiment-design​

  • Logical: data-experiment-design
  • Description: Experiment plan: hypothesis, success metric, boundaries, ramp/kill criteria, analysis checklist
  • Quick Win: No - builds on metric-spec
  • Key Points:
    • Hypothesis must be falsifiable
    • Predefine success/guardrails and decision rules
    • Track novelty effects and segment impacts
  • → Next: data-analysis-plan
  • MCP Required: GA4 (baseline + segments)
  • Input: [hypothesis]
  • Output: $JAAN_OUTPUTS_DIR/data/experiments/{slug}/experiment-design.md

/jaan-to:data-analysis-plan​

  • Logical: data-analysis-plan
  • Description: Sample sizing notes (assumptions), decision rules (ship/iterate/stop), bias + data quality checks
  • Quick Win: Yes
  • Key Points:
    • Hypothesis must be falsifiable
    • Predefine success/guardrails and decision rules
    • Track novelty effects and segment impacts
  • → Next: data-postlaunch-report
  • MCP Required: None
  • Input: [experiment]
  • Output: $JAAN_OUTPUTS_DIR/data/experiments/{slug}/analysis-plan.md

/jaan-to:data-cohort-analyze​

  • Logical: data-cohort-analyze
  • Description: Cohort/retention analysis with retention curves and churn risk identification
  • Quick Win: No - needs window functions expertise
  • AI Score: 5
  • Key Points:
    • Combine cohorts + qualitative signals
    • Identify top drop-offs and root causes
    • Output a prioritized action list
  • → Next: data-postlaunch-report
  • MCP Required: GA4 (cohort data), BigQuery (optional)
  • Input: [cohort_type] [retention_event] [periods]
  • Output: $JAAN_OUTPUTS_DIR/data/cohorts/{slug}/cohort-analysis.md
  • Failure Modes: Incomplete data; timezone issues; not accounting for seasonality
  • Quality Gates: Early cohorts stable; cross-reference with finance

/jaan-to:data-postlaunch-report​

  • Logical: data-postlaunch-report
  • Description: Insights summary + interpretation notes, chart checklist (no code), segment highlights
  • Quick Win: No - needs post-launch data
  • Key Points:
    • Combine cohorts + qualitative signals
    • Identify top drop-offs and root causes
    • Output a prioritized action list
  • → Next: pm-release-review
  • MCP Required: GA4 (post-launch data)
  • Input: [metrics]
  • Output: $JAAN_OUTPUTS_DIR/data/insights/{slug}/postlaunch-report.md

/jaan-to:data-attribution-plan​

  • Logical: data-attribution-plan
  • Description: Tracking plan + UTMs, source of truth + governance, limits/risks checklist
  • Quick Win: No - needs attribution setup
  • Key Points:
    • Attribution limits (multi-touch vs last-touch)
    • UTM hygiene and naming
    • LTV/CAC models should show assumptions
  • → Next: data-ltv-cac-model
  • MCP Required: GA4 (attribution data)
  • Input: [channels]
  • Output: $JAAN_OUTPUTS_DIR/data/growth/{slug}/attribution-plan.md

/jaan-to:data-ltv-cac-model​

  • Logical: data-ltv-cac-model
  • Description: Model inputs/outputs table, sensitivity notes (what drives outcomes), data needed to validate
  • Quick Win: Yes
  • Key Points:
    • Attribution limits (multi-touch vs last-touch)
    • UTM hygiene and naming
    • LTV/CAC models should show assumptions
  • → Next: —
  • MCP Required: None
  • Input: [assumptions]
  • Output: $JAAN_OUTPUTS_DIR/data/growth/{slug}/ltv-cac-model.md

/jaan-to:data-anomaly-triage​

  • Logical: data-anomaly-triage
  • Description: Triage pack: scope, likely causes, next checks, who to pull in, RCA starter template
  • Quick Win: No - needs multiple MCPs
  • Key Points:
    • Combine cohorts + qualitative signals
    • Identify top drop-offs and root causes
    • Output a prioritized action list
  • → Next: sre-incident-runbook
  • MCP Required: GA4 (anomaly detection), Sentry, Clarity (optional)
  • Input: [kpi]
  • Output: $JAAN_OUTPUTS_DIR/data/monitoring/{slug}/anomaly-triage.md

/jaan-to:data-sql-query​

  • Logical: data-sql-query
  • Description: Ad-hoc SQL queries from natural language with results summary
  • Quick Win: Yes - natural language to SQL
  • AI Score: 5 | Rank: #2 (2nd highest-leverage task)
  • Key Points:
    • Events are verbs; properties add context
    • Ensure consistent naming + schema
    • Validate tracking with QA and dashboards
  • → Next: —
  • MCP Required: None (schema context provided)
  • Input: [question] [tables/schema]
  • Output: $JAAN_OUTPUTS_DIR/data/queries/{slug}/query.sql
  • Failure Modes: Misunderstanding question; wrong joins; incorrect filters
  • Quality Gates: Row count sanity checks; cross-reference dashboards

/jaan-to:data-dbt-model​

  • Logical: data-dbt-model
  • Description: dbt staging/mart models with tests, documentation (schema.yml)
  • Quick Win: No - needs dbt knowledge
  • AI Score: 5 | Rank: #19
  • Key Points:
    • Events are verbs; properties add context
    • Ensure consistent naming + schema
    • Validate tracking with QA and dashboards
  • → Next: data-sql-query
  • MCP Required: dbt Cloud (optional), BigQuery/Snowflake (schema)
  • Input: [source_table] [model_type]
  • Output: $JAAN_OUTPUTS_DIR/data/dbt/{slug}/model.sql
  • Failure Modes: Circular dependencies; missing tests; poor documentation
  • Quality Gates: dbt test passes; row counts match; code review